Expert system shells in chemistry: CHIRULE, a chiral chromatographic column selection system using similarity searching and personal construct theory

Abstract

The current process of selecting a column to perform a chiral separation can be characterized as more of an art than a science. This dissertation describes CHIRULE, a "chromatographic assistant" to aid in developing chiral separations.

CHIRULE constructs an n-dimensional information space from a large number of known chiral separations by fragmenting the molecules at their chiral center, producing four molecular fragments. Molecular properties are calculated for each of these fragments. The properties are the axes used to place known separations into the information space.

To suggest a column, the target molecule is added to the information space. Similarity property searching is used to select all known separations similar to the target molecule. The chemistry-based expert system shell CHESS used to develop CHIRULE also includes features such as functional group recognition and automated knowledge extraction techniques based on Personal Construct Theory.

The results suggest that fragment property values are a route to enhanced understanding and improved selection of chiral separation methods.